Cognitive Bias in Forensic Investigation

• Recognise confirmation, contextual and expectation effects.
• Separate task-relevant information from potentially biasing narrative.
• Explain the purposes and limits of documentation, staged information and independent review.

Confirmation bias can favour information consistent with an initial explanation while contrary information receives less attention. Contextual influences arise when information beyond the material being examined affects a judgement. An initial suggestion can also anchor the range of possibilities considered.

At the scene, an early narrative may influence which locations are searched or which items are submitted. In comparison work, knowledge of an expected result can influence the interpretation of ambiguous features. Later review can repeat the same assumption if the reviewer receives the conclusion before evaluating the underlying material.

Bias risk is not proof that a particular result is wrong. A disagreement between examiners also does not, by itself, identify its cause. The observations, method, information exposure and decision record need examination.

Bias can affect the evidence set before laboratory interpretation begins. An investigator who collects only items consistent with one account may leave later examiners with an incomplete picture. A scientifically sound examination of submitted items cannot recover information from locations that were never documented or material that was discarded.

Use an explicit examination plan and record why material was selected, omitted or deferred. Safety, access and sample limitations can justify incomplete coverage. They should be visible so that a later interpretation does not assume a complete search.

An account may be revised as evidence accumulates. The important distinction is between evidence-led revision and selectively redefining the task so that every outcome appears to confirm the initial explanation.

A specimen’s origin, collection conditions and known hazards may be necessary to choose a valid method and handle it safely. Prior handling and the proposed activities may be needed for interpretation. By contrast, an allegation that a suspect “must be guilty” does not establish a scientific property of a trace.

Information relevance can change with the task. A fact unnecessary for initial feature documentation may become essential when evaluating a transfer explanation. Specify the task and reasoning rather than apply a universal rule that all context is helpful or all context is harmful.

Information management begins with a defined task. Initial documentation of an ambiguous trace may need less case narrative than an activity-level assessment. Method selection and safety may require information immediately. Do not delay essential hazard information in an attempt to make work appear blind.

Where the service uses a case manager or designated information filter, define that role and its limits. The person should understand the examination well enough to preserve necessary context. An undocumented informal message can undermine a planned exposure sequence even when the main file was carefully restricted.

Record what information was supplied and when, to the extent required by the procedure. A later claim that the examiner was unaware of a fact should be consistent with the actual communications and records.

A staged process can preserve an initial record of features before potentially influential comparison information is introduced. This makes it possible to examine whether, and why, the interpretation changed. The procedure must fit the discipline and task rather than be improvised to force every examination into one sequence.

Linear sequential unmasking is a published approach to ordering information exposure. It should not be presented as a guarantee of error-free work or as permission to suppress contrary evidence. Its implementation requires a defined procedure and an appropriate record of initial observations and revisions.

A practical exposure record can link an initial assessment, the information subsequently introduced, the date or stage of access and any resulting revision. It should show the basis of a change rather than merely state that a new conclusion was reached.

Initial decisions can be uncertain and legitimately change. A small, tentative observation is different from a confidently documented feature that is later removed after viewing a reference. Neither situation should be judged solely from the direction of the change. Examine the material, reason and procedure.

Do not turn the initial record into a prohibition on correction. The aim is to preserve the history of reasoning and make changes reviewable. If a feature was recorded in error, correct it transparently and consider any effect on the conclusion or earlier report.

Blind procedures and staged exposure have different practical limits across disciplines. A scene examiner necessarily sees the location; a pathologist may need clinical information; an activity evaluation may require case circumstances. Describe the safeguards that were feasible rather than claiming a single universal protocol.

An instrument-generated value does not eliminate every judgement. Decisions about sampling, data quality, thresholds, model settings and interpretation can remain sensitive to expectations. Validation and appropriate automation can support consistency, but their limits and the human decisions around them still need attention.

Training helps people recognise risks, but recognition alone does not demonstrate that a risk was controlled. Build the relevant checks into the workflow and examine whether they operate as intended. Avoid numerical claims about bias reduction unless a suitable study supports them.

A reviewer who knows the first conclusion may be influenced by it. Where suitable, an independent assessment before disclosure of that conclusion can reveal disagreements that would otherwise remain hidden. Record what the reviewer saw and the task they performed.

Independent review is not the same as an automatic second signature. A documentation check, a technical review and a fresh interpretation address different risks. Identify which review occurred and how disagreements were handled without deleting the earlier reasoning.

Procedures, competency assessment, workload management and suitable working conditions also matter. Telling an examiner to “be objective” is weaker than a process that controls avoidable exposure and preserves an auditable record.

An independent assessment is made before the reviewer is exposed to the first conclusion, under a defined procedure. Discussion can follow once both assessments are recorded. This distinguishes the value of a second assessment from the value of a joint technical discussion.

A reviewer may identify a missed limitation, interpret an ambiguous feature differently or agree for a different reason. Those differences can be informative even when the final conclusion remains the same. Preserve the substantive issue and how it was resolved rather than reducing the record to “checked.”

Agreement is not independent confirmation if the process required the reviewer to adopt the first answer. Equally, disagreement is not evidence that one examiner acted dishonestly. A competent review process should allow documented uncertainty and unresolved differences when justified.

An examiner should be able to explain which information was available, what process was followed and how limitations affected the conclusion. Do not describe work as fully blind if the decision-maker had relevant case knowledge, or as unbiased merely because a checklist was completed.

NIST’s human-factors work treats decision quality as a systems issue. That supports examining the interaction between people, information and procedures rather than attributing every error to an individual’s lack of integrity.

Scientific Reasoning in Crime Scene Investigation. Evidence Interpretation: Source and Activity Questions.

• Bias can affect honest and experienced practitioners.
• Manage information by task and stage, not through blanket secrecy.
• Preserve observations, exposure history and justified revisions.
• Describe review and safeguards accurately without promising zero error.